BTW, one aspect of this proposal that needs to be discussed is that
it can break existing SQL. An example discussed nearby[1] is
regression=# select * from (values (1,7), (3,4) order by "*VALUES*".column2);
column1 | column2
---------+---------
3 | 4
1 | 7
(2 rows)
We concluded in the other thread that we didn't want to risk breaking
such code. I concede that this example isn't too compelling on its
own, but I wonder if there's other cases that are more likely to be
in common use.
(If we do decide it's okay to break this, my opinion about what to
do in the other thread would change.)
regards, tom lane
[1] https://www.postgresql.org/message-id/251197.1730222362%40sss.pgh.pa.us
